xAI Launches Grok 4.5: Coding and Agent Focus to Rival Opus

On July 9, xAI officially launched the Grok 4.5 model. Described as the company's first model specifically trained for coding and agentic tasks, it was developed in collaboration with Cursor. The model is designed for real-world engineering tasks, excelling in large codebases, long multi-repository tasks, and multi-tool orchestration. Elon Musk stated that internal evaluations show Grok 4.5's capabilities are comparable to Claude Opus 4.7, but with faster speeds, lower costs, and maximum intelligence per unit of time and cost.

Core Specs and Pricing

Grok 4.5 features a 500K token context window, speeds of 80 tokens/s, and supports tool calling, structured outputs, and vision. The API is priced at $2 per million input tokens and $6 per million output tokens. Furthermore, the model is highly token-efficient; on the SWE Bench Pro, its average output token usage (15,954) was significantly lower than Claude Opus 4.8 max (67,020), saving approximately 4.2x.

Platform Availability

Grok 4.5 is now live and rolling out across multiple platforms. Users can access it via grok.com, Grok Build (requiring an update to version 0.2.92), Cursor, Hermes Agent, OpenClaw, and the xAI API, as well as gateways like OpenRouter. However, the rollout for the EU region is expected in mid-July.

2026-07-09 ~ 2026-07-10 · 55 related posts

Full story(20 episodes)→

3 near-duplicate retellings: DimaZeniuk · AGI Hunt · Kyrannio